How hard it is to build an AI app between riders and drivers which will autoscale itself based on network bandwidth.... use blockchain to store data on users mobile.... and connects riders to driver based on distance and just send sms to rider of driver number ... completely automated runs by itself and doesnt need any developers to run it.... basically AI driven decentralized bot ....

"AI", "bot", "eth", "autoscaling", "blockchain", "decentralized" Do you have a clear idea of what these terms mean?

AI : a system which will take decisions on its own Bot : which can interact with users Autoscaling : automarically add or remove nodes based on traffic on network Blockchain : to store ledger on multiple computers ( mobile ) to make sure of no tampering of data . No need to store data on pods as we need stateless application to scale better Decentralized : every users mobile can act as data node

How do you plan to handle malicious nodes who will deliberately poison with fake rider/driver data? (and no, this is not fiction, Uber and Ola actually did this to try create a bad experience for their competition)

Lets not add complexity at this stage ... lets assume all users ( riders and drivers ) are loyal .... first we need to make sure if simple use case can work before start adding layers of complexity .... also data is very simple just number and location and output is sms with number ....

What's the goal of this activity? Learning stuff? sure seems fun. Making something practically useful? Let us identify the already solved bits and the unsolved problems. (IMO, lack of tech is NOT the unsolved problem)

This has to start as a fun .... we just need 4 experts from each domain : data ( i can take lead ) , blockchain , kubernetes , UI app developer .....

seems like throwing a bunch of buzzwords for no real reason. autoscaling & k8s for 5 nodes/10 riders/few kbs of data is just wasteful, as is 'blockchain'

Ahh its not about optimization here .... more about if all these things can work together or not coherently for a very basic use case ......

Not very knowledgeable but this looks a lot like P2P networking with synced state. You don't really need any AI or blockchain

No you need blockchain to reduce cost of traffic ( and keep subscription cost to end user minimal ) and store data in tamper proof manner .... use users mobile harddisk to store data ....

peak buzzwordify content

The secret to making it work is to paint the Bitcoin blue.

For those who are interested : Assumptions : Drivers and passengers dont change. ( no new addition or deletion ) Only their location changes . They could be active or inactive No one cheats Biggest goal of this exercise is to reduce cost of transferring signal from passenger to driver and vice versa . Data part : Take an example of Tier 2 city in india Radius of this city is 25km Number of drivers : 25k Number of passengers every day : 2.5 lakhs Data of drivers : 25k x 1kb = 25mbytes Each User can easily store this data on their mobile hard disk . This means all users( driver and pasengers ) have all data of drivers . This completely eliminates heavy lift and shift of data . Pasengers data : 2.5 lkhs x 1kb = 2.5gb Here you can give choice to driver to store all data or store it in shards ( shards based on pincodes ) ============================ We have eliminated major data transfer part by storing all drivers data locally . Now lets assume all users in network has all data of drivers but drivers have data only about their most preferred pincodes . Simple use case : User PA wants to go from location X1 to Z1 . PA sends singal to centralized node : CS1 Driver : DA DB DC all three also sends signal to CS1 about their availability . CS1 checks if thers any P$ available and sends id of DB to PA as he is closest one . Data of DB ( his number name and other details ) already exists on PAs harddisk which then matched by local part of program and autheticated with blkchain tech ( on mobile ) and it sends sms to PA. Once accepted signal goes back to central node and he can flag both PA and DB Data transfer cost : only ID back and forth. To eliminate traffic congestion of signals ( only ids ) : Now think about cengralized nodes network is very similar to ring roads in city. Inner then outer and so on .... and you have chowks at equidistant post on each of these ring roads circumference . So concept of central node is the closest node on closest ringroad circumfernece.

Aren't you optimising for the wrong thing? Data is almost free - both network transfer and local storage for a few GB. 🤓 If on the other hand, the goal is to have the database stored in a distributed network w/o resorting to a company (Uber, Ola), then you've defeated the design by having a central node. Uber isn't selling you a taxi ride - it is selling you a predictable experience wherever you go (ala McDonalds, Subway) - clean cars, trained drivers, no price haggling or overcharging, predictable arrival. You could build such a network based purely on reviews but the amount of bad actors (spam) could easily overwhelm your "AI" into wrong decisions. Enforcing quality requires a few humans, I'm afraid. Interesting thought experiment though.

Architecture will like this .... idea of central node is also decentralized

message has been deleted

Ok since you've continued this the next day, you're most likely not drunk texting. And you seem to be absolutely serious about this, so in the spirit of this community I will genuinely try to help you once. CDN + lambda invocation is extremely cheap. Cloudflare workers with their edge KV stores might be in similar or cheaper range. i.e. you're trying to over-engineer solve one of the cheap and pretty much solved parts of this entire business. ----------------------------------------------------------------------- In all of what you're thinking about, the one thing that can actually make you shit tons of clout/money/industry defining leader is if you can set up reliable edge nodes at the scale you've drawn them to be. It's extremely hard to do well and companies have been trying to make it economically viable for a long time.

Most of the things i have talked are easy problems and have been solved but question is have this been solved in autonomous way at scale ..... goal is not to build full fledged taxi app but think of an sms service …. where i need to book cab and i receive 5 sms sorted by drivers distance …. no money involved and no ratings and review … only goal is can this scale to infinity with no human involved …..

Short answer: Not right now, not in this form for sure. Long answer: Assume everything is possible. Anything that you think of, write down the cost of making it + running it vs existing industry alternative. If it costs more, it's not worth pursuing.